On the use of tolerant graded inclusions in information retrieval

In this contribution, a fuzzy-set-based information retrieval model is considered. First, we show that the retrieval mechanism of such a retrieval model can be defined in terms of graded inclusion. This approach derives from the notion of division of fuzzy relations in the framework of Database Management Systems. Then, we point out different lines of extension of the graded inclusion aimed at making it more tolerant (to exceptions, in particular). It is shown that the use of such tolerant inclusion indicators reduces the risk of obtaining empty answers. MOTS-CLÉS : Modèle flou de recherche d’information, inclusion graduelle, inclusion tolérante, relations floues, termes d’index pondérés, requêtes pondérées.